Mac SE Radius '020 25mhz Benchmark Results

[JeffC]

I thought someone here might be interested in some benchmark results from my accelerated SE. The machine has a Radius 25mhz 68020 accelerator with 68881 FPU. 4mb RAM, spinning-platter HD. The SE/30 used in the tests has 8mb RAM, Mac SD drive, and a Radius 40mhz 68030 accelerator. Both accelerated machines were run with the appropriate accelerator drivers and cache enabled, except for the last test which compared the '020 SE with the cache and math extension against the same machine without the extensions. Both machines were running System 6 (I think 6.0.8, though one might have been running 6.0.7). The tests were run with Speedometer 3.06. I benchmarked the stock SE, which not surprisingly was a 1.0 on most tests, which lines up with the Mac Classic that Speedometer 3.06 uses as a baseline. I'm not sure why the disk speed was so slow for the stock SE, but I'm concerned since it seems fine with the accelerator enabled.

 

[retr01]

Hi @JeffC!

Wow! Thank you for sharing the benchmarks.

The stock SE CPU is part of the limiting factor for the SCSI. Have you run a benchmark for the hard disk drive focused without all other measurements?

I have a question. Does the 25 MHz Radius accelerator in the SE up the memory limit from 4 MB to higher? How many address lines are connected out of the 68020?

 

[JeffC]

Hi @JeffC!

Wow! Thank you for sharing the benchmarks.

The stock SE CPU is part of the limiting factor for the SCSI. Have you run a benchmark for the hard disk drive focused without all other measurements?

I have a question. Does the 25 MHz Radius accelerator in the SE up the memory limit from 4 MB to higher? How many address lines are connected out of the 68020?

No I haven't done any disk-specific benchmarks. The "stock" SE in the comparison was the accelerated SE with the accelerator disabled, I might try running the same benchmark on one of my non-accelerated SEs and see if I get the same low results. Maybe the low disk score had to do with the accelerator being installed, even though it was disabled? Since the disk benchmark seems fine when the accelerator is enabled, I didn't put much thought into it, I was more interested in the processor performance.

The Radius accelerators for the SE (16 and 25mhz) don't have additional RAM slots on the accelerator board, and I'm 99.9% sure they don't do anything that would allow more than 4mb on the main board. This creates a performance bottleneck, which Radius tried to mitigate using a 32k cache. There is an accelerator review article in the March 1989 issue of Mac World (https://vintageapple.org/macworld/pdf/MacWorld_8903_March_1989.pdf), pages 119-127. On page 124 there is a paragraph describing the Radius cache. These are the magazine page numbers, they may not correspond to the PDF page numbers. I'm not certain on your question about address lines, I'm sure someone on here can probably answer that better than I can. Is your question about the 68020 in general, or this board in specific?

 

[retr01]

I'm not certain on your question about address lines, I'm sure someone on here can probably answer that better than I can. Is your question about the 68020 in general, or this board in specific?

I have been musing about this for a while and still am. The Motorola 68000 is a 32-bit Macintosh Plus and SE stock CPU that can address from 0 to 23 lines connecting to the logic board. That means it has 24-bit addressing. Since the 68000 addresses 24 bits or lines, 2 to the 24th power equals 16,777,216 bytes. Hence, the 68000 can address a maximum of 16,384 kilobytes or 16 megabytes of RAM.

The problem is that Apple only connected 22 out of 24 lines from the 68000 to the logic board. That is an odd and haphazard 22-bit addressing, not the standard 24-bit addressing. That severely limited the maximum addressable memory to 2 to the 22nd power, which is 4,194,304 bytes or 4,096 kilobytes, equal to 4 MB. Apple further murked the waters by imposing mediocre memory mapping in ROM and the System software's memory manager.

I do not understand why we cannot connect two more lines from the 68000 on the SE Reloaded board in the Gerber file, which would increase the memory ceiling for the SE. It still does not make sense why we are putting up with the 4 MB limit and bowing to Apple's stupidity back in the day. Why are we repeating history?

I love the Plus and SE for it it is, but I think it can have 16 MB of RAM stock with the right modern tweaks.